A diagnosis of the presence of a lesion or the kind of the lesion, so called a pathological diagnosis, has been actively performed by observing a tissue slice obtained from a human or an animal with a microscope. In the pathological diagnosis, the obtained tissue sample is subjected to dehydration process for fixation, blocking process with paraffin, slicing process to cut the tissue sample into thin slices having a thickness of 2 to 8 μm, removing process of the paraffin, staining process, and observation process with a microscope to obtain image data (cell image). The diagnosis is includes analysis of change in size and shape of a cell nucleus, morphological information such as tissue pattern change, and staining information. When extraction of region of interest (for example, cell region) from a cell image is performed manually in the image processing, it takes a lot of time and causes large errors due to the difference of the operator, for example.
Recently, many techniques of automatic image analysis are suggested for the purpose of grasping cancer region efficiently, in which a specific protein is overexpressed, from an image of the tissue slice. For example, according to one of the suggested techniques, outline of individual cells are automatically extracted even when plural cells overlap in an image of a sample.
For example, according to the technique described in Patent Document 1, the shapes of individual cells can be extracted even when plural cells overlap with each other in the cell image (see paragraph 0018). In this technique, specifically by focusing on the dyeing concentration (concentration gradient) of cells, the shapes of individual cells are intended to be extracted on the basis of positive or negative codes of an inner product value of a concentration gradient vector at a pixel constituting the outline of a cell and a displacement vector from the pixel to the center of the cell (see paragraphs 0027 to 0028, FIG. 10, paragraphs 0084 to 0088, FIGS. 13 to 16, and the like).
According to the technique described in Patent Document 2, a constriction of the outline of overlapping plural cells is recognized as a boundary of individual cells using Watershed method, and each of the regions divided by the boundary is detected as a single cell.
According to the technique described in Patent Document 3, the region of interest in an image is automatically recognized by evaluating attribute values of each pixel in the image on the basis of feature amount of the observation target in a cell.